7640 & 7643

[dschief]

picked up these 2 today; system is showing that it will take 25 hr 50 min to complete.

have never had a gpu wu take more than 5 hr.

both are running in the same box on a pair of GTX460 se's. does 25+ hrs. seem reasonable?

[7im]

WUs always get bigger, and that mantra has been repeated, repeatedly.

Yes, very reasonable, also noting the PG announcement about these projects going main stream.

Also, I don't think this is a V7 related issue.

See also http://foldingforum.org/viewtopic.php?f ... 77#p211806

[Hyperlife]

As a comparison point, my stock GTX 460 (non-SE) has a 12:26 TPF for p7643. It takes almost 21 hours to complete one.

[sswilson]

Yeah, my 560 ti cards are showing 10+ minutes TPF if it's any consolation.

PPD on these really bites as well especially when we were used to the high heat/PPD ones we just got through. Previous average was around 13K PPD, high average for the last batch was around 18K, these ones are coming in under 11K PPD.

[bruce]

Yeah, my 560 ti cards are showing 10+ minutes TPF if it's any consolation.

PPD on these really bites as well especially when we were used to the high heat/PPD ones we just got through. Previous average was around 13K PPD, high average for the last batch was around 18K, these ones are coming in under 11K PPD.

Can somebody report the same information for a GTX 460 card?

[bollix47]

On my GTX 460 running a P7643 I'm getting just under 9K PPD whereas on the P803x it was up around 13K. TPF=12:23

[bruce]

I'm getting 7143 PPD on a P8034 with lots of lag. (Win7-64 -- GTX460 non-SE) TPF=7:46. I'm not sure we can successfully generalize.

[sswilson]

These figures are taken from what's reported by the beta client.....

560 ti @ 850 core

WU 7641 (245,0,1) / TPF 10 minutes 8 seconds / PPD 10953.27
WU 7644 (122,0,2) / TPF 9 Minutes 52 seconds / PPD 112 50.99

run on a 1090t @ 3.8 (smp on 4 cores, w/ 2 cores left open for GPU folding)

560 ti @ 900 Core

WU 7642 (130,0,1) / TPF 17 Minutes 25 Seconds / PPD 6381.14

GTS 450 @ 850 core

WU 7642 (437,0,0) / TPF 21 minutes 23 seconds / PPD 5194

Run on an athlon X2 @ 3.6


The output of the third 560 ti on the Athlon dual core is majorly nerfed which makes me wonder if these WUs are greatly affected by the system proc / memory interface.

edit: All three of these 560 ti cards (on the same systems they're currently on) were getting around 18K on the previous set of WUs (the extra hot ones.... ).

[*hondo*]

Is it possible to re-evaluate these please, I reckon these are losing me up-wards of 5000 PPD against the 80XX series WUs

Although the GPU is set @ 100% load. In MSI Afterburner it's showing only 97% load fortunately though, the temps are down to 63C

Project : 7643
Core : GPU3 OpenMM
Frames : 100
Credit : 7718

Code: Select all

 
[16:36:09] Project: 7643 (Run 427, Clone 0, Gen 1)     [16:36:09] 
[16:36:09] Assembly optimizations on if available.
[16:36:09] Entering M.D.
[16:36:11] Tpr hash work/wudata_01.tpr:  387610028 1081406864 1851892430 3291987443 864452771
[16:36:11] GPU device info: vendor=0 device=0 name=<NA> match=0
[16:36:12] Working on Protein in water
[16:36:12] Client config found, loading data.
[16:36:12] Starting GUI Server
[16:37:39] Setting checkpoint frequency: 25000
[16:37:39] Completed         3 out of 2500000 steps (0%).
[16:43:46] Completed     25000 out of 2500000 steps (1%).
[16:49:59] Completed     50000 out of 2500000 steps (2%).
[16:56:08] Completed     75000 out of 2500000 steps (3%).
[17:02:22] Completed    100000 out of 2500000 steps (4%).
[17:08:38] Completed    125000 out of 2500000 steps (5%).
[17:14:47] Completed    150000 out of 2500000 steps (6%).
[17:20:55] Completed    175000 out of 2500000 steps (7%).
[17:27:06] Completed    200000 out of 2500000 steps (8%).
[17:33:21] Completed    225000 out of 2500000 steps (9%).
[17:39:31] Completed    250000 out of 2500000 steps (10%).
[17:45:42] Completed    275000 out of 2500000 steps (11%).
[17:51:52] Completed    300000 out of 2500000 steps (12%).
[17:58:03] Completed    325000 out of 2500000 steps (13%).
[18:04:15] Completed    350000 out of 2500000 steps (14%).
[18:10:27] Completed    375000 out of 2500000 steps (15%).
[18:16:39] Completed    400000 out of 2500000 steps (16%).
[18:22:51] Completed    425000 out of 2500000 steps (17%).
[18:29:03] Completed    450000 out of 2500000 steps (18%).
[18:35:17] Completed    475000 out of 2500000 steps (19%).
[18:41:30] Completed    500000 out of 2500000 steps (20%).

[guest3412]

I know that I'm not running one of those two projects, but I got a 7642 today also 21443 atoms, and it says that my tpf is 9mins and it'll take 15hrs to complete on my GTX 480

Project Summary is giving 13 days for preferred deadline and 7718 points

I'm also not getting 100% GPU usage, hovering at 70~80%

I would think that if the projects were adjusted slightly - the GPU could be better utilized (100%).

[P5-133XL]

I was getting 7,800 PPD on GTX 460's (800MHz - 825MHz clocks) running p764x; I get 11,000+ running p762x; I get 14,000+ running p803x. This seems to be relatively consistent across multiple cards running on multiple machines.

[*hondo*]

This is how the current WU scoring system seems to be working here.

On the assumption of a static bench mark PC, I’d be very interested in knowing just how the differences stack up?

Project 7643 in 10.566 hours = 7718 Points

Versus 80xx in 4.02 Hrs = 3843 Points

80xx in 4.02 Hrs = 3843 Points X 2

Therefore I can Fold 2 X 80xx = 7686 Points

With 2.57 Hours to spare

Come on Stanford please get the PPD right

[HaloJones]

I have only heard people complaining about these. What card were they benchmarked on that the ppd stayed the same as the 80xx work units?

[sswilson]

I believe Stanford has typically applied a "30%" possible range on what PPD might be expected, and believe it or not, these come close to being 30% below recent standards. What's making them look soooooo bad is the fact that we just got through a batch of WUs which were providing PPD well over 30% above what has been the standard.

I'd like to see a more consistant PPD being applied, but am willing to cut Stanford some slack on this one.

[*hondo*]

Is that sort of like a 60% divide? I guess I've been participating in F@H since 2003-4ish and am still unsure where the middle line is.

I certainly wouldn't expect a computer to allow any deviation when using Excel to keep tabbs on my incoming and out goings.